Explaining Unfairness in GNN-based Recommendation

Ludovico Boratto, Francesco Fabbri, Gianni Fenu, Mirko Marras, Giacomo Medda

Keywords: Recommender Systems, User Fairness, Explanation, Graph Neural Networks, Counterfactual Reasoning

Abstract: Nowadays, research into personalization has been focusing on explainability and fairness. Several approaches are able to explain individual recommendations, but unfairness explainability has been limited to finding user/item features mostly related to biased recommendations. In this paper, we devised a novel algorithm that leverages counterfactuality methods to discover user-item interactions as user unfairness explanations in GNN-based recommendation. Our approach perturbs the graph topological structure to find an altered version (counterfactual explanation) that minimizes the disparity in utility between the protected and unprotected demographic groups. Experiments on four real-world graphs showed that our method can systematically explain user unfairness on three state-of-the-art GNN-based recommendation models. Moreover, the generated unfairness explanations are connected to properties related to the networks topological aspects. Source code and datasets: https://github.com/jackmedda/RS-BGExplainer.

Submission Type: Extended abstract (max 4 main pages).

Agreement: Check this if you are okay with being contacted to participate in an anonymous survey.

Submission Number: 66